Conditional Visuomotor Learning and Viability Theory

Abstract

In conditional visuomotor learning, several arbitrary associations between visual cues and motor responses have to be learned by trial and error at the same time. Monkeys, as humans, do not achieve this task by randomly trying each possible association. Rather, they use a strategy that organizes sequentially the acquisition of individual stimulus-response associations. Accordingly, neuronal recordings in the monkey striatum, the main basal ganglia structure, reveals two forms of plasticity during learning, a transient one that could constitute the neuronal correlate of the strategy, and a long-lasting one that could reflect the slow neuronal implementation of individual associations. Existing models of basal ganglia function based on reinforcement learning cannot account for this dual process. Hence, we developed a mathematical model of conditional visuomotor learning, inspired from viability theory, which implements both the formation of individual associations and the use of strategies to organize learning.